Jensen Huang, co-founder and chief executive of Nvidia, has cautioned British companies, government agencies, and individual users against relying on “opaque” artificial intelligence models developed by major American labs for critical functions. Speaking in London on Wednesday, Huang urged the UK to prioritize open-source AI models, which he said provide greater transparency and control over data—a crucial factor for mission-critical applications where trust, regulation, sovereignty, and proprietary concerns are paramount.

According to Huang, proprietary AI systems such as those from OpenAI and Anthropic require users to send data to external servers and offer limited opportunities for customization. In contrast, open-source models from companies like Meta, Mistral, and various Chinese firms can be downloaded and fine-tuned on-site, allowing organisations to retain data internally and better tailor AI tools to their specific needs. This has drawn increased attention from governments and regulated sectors advocating for sovereign AI solutions.

Huang acknowledged that opaque AI services remain suitable for non-essential tasks where full transparency is less critical. However, he emphasized that relying exclusively on closed systems for sensitive areas is not sustainable. “How could you have an opaque agentic system contained in an opaque box, run by an opaque company?” he asked.

At the London event, Huang met with politicians, including Andy Burnham, whom he told that AI presents an immediate economic opportunity for the UK, describing the country as being at a “Goldilocks moment” with the necessary intellectual capital to lead in the emerging AI industry. Huang highlighted Britain’s historical contributions to computing and the internet and suggested its geopolitical position enables it to mediate AI-related dialogue between the US and China.

Despite Nvidia’s dominance in AI chip supply, Huang noted that major tech companies like OpenAI and Google are developing their own chips, which may reduce their reliance on Nvidia hardware. Still, he emphasized that both proprietary and open approaches have roles to play in the AI ecosystem.

Huang also addressed concerns about AI’s risks, describing fears about existential threats as lacking scientific basis and characterizing safety challenges as issues of technology and engineering rather than regulation. He compared AI safety to engineering safer cars, stressing the importance of containment, monitoring, and technological solutions over new laws.

Continuing Nvidia’s commitment to the UK AI sector, the company announced investments totaling $1.16 billion, including minority stakes in five British start-ups, and pledged computing resources worth $250 million to AI safety research institutes based in the country. Huang’s visit also included meetings with start-ups and policymakers ahead of Nvidia’s annual showcase in Berlin, underscoring the firm’s ongoing engagement with the UK technology landscape.

While Huang praised the UK’s potential, he urged the country to cultivate greater entrepreneurial spirit, noting a cultural tendency to focus on barriers to success rather than opportunities, which he said could hinder innovation in the AI field.